To all those who say you can't win using speed & pace figures only...

[PaceAdvantage]

Originally Posted by traynor

I think it is unfortunate that DeltaLover sometimes tends to underemphasize points that are more important/significant than the casual reader might believe. An example is emphasizing actual return over "ROI" (two VERY different concepts that many seem to equate).

The ROI on a specific set of races is overfit to that specific set of races. There is ZERO guarantee (or even indication) that the specific set of races used represents a normal distribution over a larger set of races. Using ROI as a criteria leads one to chase rainbows and lose money. That is not a trivial matter, nor is it a spurious statement made for the sake of provoking argumentation. It matters not at all what the "ROI" of a given set of races was or was not. It only matters how much money you gained or lost by wagering on that particular set of races.

Not to be rude or anything, but once again, the "smartest guy in the room" steps in and basically says nothing.

What exactly is the point of the above? Any measurement in the world is OVERFIT to what is being measured....because it's the measure of what is being measured!

That doesn't make it useless, of course, or we wouldn't be wasting our time measuring in the first place.

ZERO INDICATION? Of course ROI over a large enough number of races IS an INDICATION of something. If your ROI over 1000 races is 0.79 per 1.00, it's a huge indication that your method used over those races sucks.

[PaceAdvantage]

Originally Posted by traynor

Consider--if the additional 100 race sample had included an outlier (a larger than normal mutuel) it might have tipped the ROI of the entire set to increase, rather than decrease. Would that have "proven" anything beyond the results of that specific strategy in that specific set of races? Not really. Would an additional 100 race set indicate continued decline in ROI, or reverse the "trend" and show an overall increase? And would either event "prove" anything beyond the results of that specific strategy in that specific set of races? I don't think so.

Again, that is not a casual observation or spurious opinion. The app I use generates dozens of "positive ROI wagering strategies" routinely. Most are worthless--which can readily be determined by applying those strategies to different sets of races. Believing that strategies developed by study of a specific group of races can be universally applied to other races with identical result can be VERY costly.

BTW, there were zero outliers in my 500 race test. In fact, I can't recall any horse that paid much over $20 during my entire run...certainly none that paid over $40.

[DeltaLover]

Originally Posted by traynor

I think it is unfortunate that DeltaLover sometimes tends to underemphasize points that are more important/significant than the casual reader might believe. An example is emphasizing actual return over "ROI" (two VERY different concepts that many seem to equate).

The ROI on a specific set of races is overfit to that specific set of races. There is ZERO guarantee (or even indication) that the specific set of races used represents a normal distribution over a larger set of races. Using ROI as a criteria leads one to chase rainbows and lose money. That is not a trivial matter, nor is it a spurious statement made for the sake of provoking argumentation. It matters not at all what the "ROI" of a given set of races was or was not. It only matters how much money you gained or lost by wagering on that particular set of races.

You are obviously right and whoever tries to disagree with what you are stating here, needs to improve his understanding of the game! (I do not think that I have said anything different in this or in any other thread!)

[traynor]

I think Lawrence Revere's comment, "It doesn't matter who wins or who loses. All that matters is who ends up with the money." applies even more to horse race wagering than it does to blackjack.

[ReplayRandall]

Originally Posted by traynor

It matters not at all what the "ROI" of a given set of races was or was not. It only matters how much money you gained or lost by wagering on that particular set of races.

You must be implementing a gradual betting scale for your "sets of races", in order to state that ROI doesn't matter....

[traynor]

Originally Posted by PaceAdvantage

Not to be rude or anything, but once again, the "smartest guy in the room" steps in and basically says nothing.

What exactly is the point of the above? Any measurement in the world is OVERFIT to what is being measured....because it's the measure of what is being measured!

That doesn't make it useless, of course, or we wouldn't be wasting our time measuring in the first place.

ZERO INDICATION? Of course ROI over a large enough number of races IS an INDICATION of something. If your ROI over 1000 races is 0.79 per 1.00, it's a huge indication that your method used over those races sucks.

If the ROI was actual, and not just on paper, it certainly does. Studying a group of 1000 races and finding a paper ROI of 1.05 does not in any way guarantee a 5% profit from identical wagers made on the next 1000 races. Actual return is meaningful--paper ROI is much less so.

[traynor]

Originally Posted by ReplayRandall

You must be implementing a gradual betting scale for your "sets of races", in order to state that ROI doesn't matter....

Actual ROI matters. Paper ROI not so much. The study of the races on this thread is impressive, because it was made "going forward" (rather than retroactively determined by parsing past races). The additional 100 races added to the initial 400 indicated a diminished ROI. It could just as easily have gone the other way.

[thaskalos]

Originally Posted by traynor

If the ROI was actual, and not just on paper, it certainly does. Studying a group of 1000 races and finding a paper ROI of 1.05 does not in any way guarantee a 5% profit from identical wagers made on the next 1000 races. Actual return is meaningful--paper ROI is much less so.

Yes...the next 1000 races will surely garner a different ROI. But they will generate a different "actual return", as well. Whether the bettor uses the ROI or the "actual return" as a measuring stick...the next cycle of bets will always be different than the preceding cycle of bets.

[DeltaLover]

Originally Posted by PaceAdvantage

BTW, there were zero outliers in my 500 race test. In fact, I can't recall any horse that paid much over $20 during my entire run...certainly none that paid over $40.

It is not that much of individual outliers but more about random anomalies that materialize across the sample that might hurt the convexity of your curve with noise.

As an analogy, you can think of coin toss game without any takeout at all. Assume that you have 100 players each of them believing in some kind of a fallacy. For example one player might believe that a running streak of 10 will follow a 1-2-1 pattern or anything like this. Playing this game for a long sequence of repetitions will end up with approximately the half of the players winners and the other half losers. Good luck trying to convince some of the winners that their “system” is simply a fallacy and a random anomaly rather than a systematic advantage their “strategy” has.

[PaceAdvantage]

Originally Posted by traynor

If the ROI was actual, and not just on paper, it certainly does. Studying a group of 1000 races and finding a paper ROI of 1.05 does not in any way guarantee a 5% profit from identical wagers made on the next 1000 races. Actual return is meaningful--paper ROI is much less so.

I bet on almost every race I posted...these were all live plays...nothing backfitted...this was a live test from the start.

In fact, the method of play I used to post all of these plays was never conceived in any way by backfitting anything in a database on past races.

[traynor]

Originally Posted by thaskalos

Yes...the next 1000 races will surely garner a different ROI. But they will generate a different "actual return", as well. Whether the bettor uses the ROI or the "actual return" as a measuring stick...the next cycle of bets will always be different than the preceding cycle of bets.

Absolutely. That was the point. Relying on a paper ROI derived from studying a specific set of races and believing that value will be perfectly replicated in another specific set of races is a misinterpretation of (the meaning of) data that can be costly.

[thaskalos]

Originally Posted by traynor

Absolutely. That was the point. Relying on a paper ROI derived from studying a specific set of races and believing that value will be perfectly replicated in another specific set of races is a misinterpretation of (the meaning of) data that can be costly.

I am in general agreement with you...but I can't understand the distinction that you are making between "actual ROI", and "paper ROI". I had a stock broker once...who kept assuring me that my stock losses were only "paper losses". But these "paper losses" always represented ACTUAL losses, to me.

[traynor]

Originally Posted by PaceAdvantage

I bet on almost every race I posted...these were all live plays...nothing backfitted...this was a live test from the start.

In fact, the method of play I used to post all of these plays was never conceived in any way by backfitting anything in a database on past races.

I understand (and fully appreciate) that. However, as thaskalos mentioned, another (equivalent) set of races could result in a different (and possibly significantly different) outcome. And, as I have mentioned several times, that outcome could just as easily be better than the outcome of the original 400 race sample as it could be worse, or just the same.

[PaceAdvantage]

So are you trying to say you have a method that allows you to test/measure/whatever, and is virtually guaranteed to perform similarly going forward, however many number of races?

If not, then I think we're all in agreement that no measure of the past, in any way, shape, or form, is guaranteed to be duplicated (or even nearly duplicated) in the future when it comes to racing.

[traynor]

Originally Posted by thaskalos

I am in general agreement with you...but I can't understand the distinction that you are making between "actual ROI", and "paper ROI". I had a stock broker once...who was assuring me that my stock losses were only "paper losses". But these "paper losses" always represented ACTUAL losses, to me.

Paper ROI is equvialent to Ray Taulbot (and crew) or many of the old PRN "systems" developing methods based on small samples. Based on results not wagered (only studied), rather than on actual betting. "Actual ROI" is the bottom line result of wagers made--not a calculation of woulda coulda shoulda that "indicates a profit."

Some do not seem to distinguish between "ROI" as a value calculated in "analysis" of past races and money-in-hand from wagering on those same races. Both are referred to as "ROI."